73 results about "Exposure compensation" patented technology

The invention discloses a three-dimensional scene reconstruction method and device based on large model geometric prior, and a medium, and aims to solve the problems that a conventional 3DGS is liable to have artifacts and detail loss in geometric discontinuity, data redundancy and illumination variation scenes, and predicts a dense depth map and a normal map from a monocular image by using a pre-trained large model. The position and form of the Gaussian kernel are constrained as additional geometric priori; a primitive adjustment strategy based on kernel density estimation is introduced in the training stage, small Gaussian primitives with similar structures and adjacent spaces are combined into a large Gaussian primitive, the rendering quality is kept, redundancy is reduced, and the volume of the model is reduced; an exposure coefficient is adaptively estimated for each input image, an exposure compensation image loss function is constructed, and floating artifacts caused by illumination differences at shooting moments are eliminated. Experiments show that compared with the prior art, the method improves the three-dimensional reconstruction precision and real-time rendering quality of complex illumination and less-texture areas in a public data set and an unmanned aerial vehicle aerial photography scene.

The invention provides an intelligent toothbrush based on AI visual identification and an oral health assessment method, and relates to the technical field of artificial intelligence and computer vision. The intelligent toothbrush based on AI visual identification and the oral health assessment method comprise the following steps: S1, based on a built-in multispectral camera of the intelligent toothbrush, dynamically adjusting illumination parameters of different oral regions by adopting a visible light and near-infrared bimodal synchronous acquisition algorithm and through a self-adaptive exposure compensation technology, high-resolution images and transmission data of tooth surfaces and gingival lines are obtained, and a multi-mode oral cavity image data set is generated. Illumination parameters are dynamically adjusted through a multispectral camera to obtain high-resolution bimodal data, the focus detection precision is improved in combination with cross-modal feature fusion, the problem that traditional single-modal imaging is insufficient in early-stage dental caries recognition is solved, and the accuracy of early-stage dental caries recognition is improved based on laser radar point cloud complementation and a three-dimensional surface reconstruction technology. And mapping the two-dimensional focus label into a dynamic three-dimensional plaque distribution model.

The invention relates to the technical field of paper marking image acquisition optimization, in particular to an intelligent paper marking image acquisition system and method based on AI and terminal equipment. An original image is acquired through an image acquisition unit, and a color temperature distribution thermodynamic diagram is generated by a light source analysis unit through a convolutional neural network; the dynamic compensation unit divides compensation grids and executes exposure compensation, the self-adaptive correction unit outputs a target image through bilinear interpolation and color space conversion, and the quality evaluation unit compares a positioning mark color gamut value to trigger a grading re-shooting mechanism. The problems of insufficient complex illumination adaptability and poor image quality stability in the prior art are solved, highlight-shadow intelligent compensation, color space adaptive conversion and quality closed-loop control of the answer sheet image are realized through AI-driven multi-unit cooperation, the full-filling point recognition accuracy and the image qualification rate are improved, and the image quality is improved. And the high-efficiency examination paper marking requirement of large-scale examination is met.

The invention provides a binocular vision global positioning system and method for a mobile robot, and relates to the technical field of robot vision navigation, and the method comprises the steps: carrying out the adaptive exposure compensation of binocular images of a left camera and a right camera, extracting edge contour feature points irrelevant to illumination, and generating a high-quality depth feature map through the combination of parallax calculation; meanwhile, feature units are constructed according to the inflection points, so that the global position of the mobile robot is determined. Therefore, the problem that the global positioning precision is reduced due to unstable visual feature extraction caused by rapid change of illumination intensity of the mobile robot in a multi-region switching scene can be solved to a certain extent.

The invention provides a vehicle management method based on visual identification. Comprising the following steps: performing visual image acquisition and basic feature identification on a vehicle in a target area, executing multi-source interference perception, acquiring raindrop volume concentration, a strong light incidence angle, vehicle tail gas heat flux density and a light intensity sudden change rate of the target area, and generating multi-factor coupling interference intensity data; executing dynamic exposure control based on the multi-factor coupling interference intensity data, outputting an exposure compensation coefficient and driving an image acquisition unit to adjust the shutter speed and the exposure duration; performing distortion correction feature extraction based on the image after exposure compensation coefficient correction, calculating image marginal definition and feature matching frequency deviation, and generating feature extraction confidence; judging whether to output a vehicle management recognition result or not according to the feature extraction confidence; the method can effectively deal with the composite interference scene, and solves the problems of low recognition precision and response lag.

The invention discloses a load balancing Gaussian sputtering method for large-scale scene geometric reconstruction. The method comprises the following steps: step 1, acquiring a multi-view RGB image sequence of a large-scale scene, and processing the multi-view RGB image sequence by using a COLMAP tool to obtain camera parameters and an initial sparse point cloud; 2, based on a pixel-level load balancing strategy, effective Gaussian primitives are screened, spatial distribution of the effective Gaussian primitives is regulated and controlled through a loss function, and redundant primitives are reduced; 3, introducing global consistent geometric prior, fusing monocular depth and sparse point cloud depth to generate a reference depth map, and constructing depth consistency and smooth loss; 4, dividing illumination areas through a multi-area exposure compensation model, adaptively correcting brightness, and constructing image luminosity loss; and 5, performing global optimization in combination with multi-dimensional loss. And 6, rendering a depth map for the trained model, generating a TSDF field by using a TSDF fusion algorithm, and extracting a scene grid to obtain a final three-dimensional geometric model. According to the method, large-scale scene high-precision reconstruction can be realized on a single GPU, memory consumption is remarkably reduced, and geometric reconstruction quality and illumination robustness are improved.

The invention relates to an image acquisition method and device, equipment, a storage medium and a program product. The method comprises the following steps: determining brightness characteristics of a preview image displayed in a viewing frame of image acquisition equipment; inputting the brightness characteristics into an exposure compensation model to obtain a target compensation proportion; performing exposure compensation on exposure parameters of the image acquisition equipment according to the target compensation proportion; and in response to an image acquisition instruction, performing image acquisition on the preview image according to the exposure parameter after exposure compensation to obtain a target image. By adopting the method, the accuracy of the exposure quantity adopted during image acquisition can be improved, so that the adaptability of the brightness of the acquired image and the ambient brightness is improved, and the image quality is further improved.

The invention discloses a photoetching machine multi-light-source cooperative high-precision exposure control system, and relates to the technical field of photoetching machine control. Initial energy parameters of all light sources are obtained, a coupling feature set of energy change and light spot morphology change is constructed, and light intensity errors, phase errors and light spot distortion errors are used as input, so that a multi-light-source cooperative high-precision exposure control system is constructed; respectively generating an inter-light-source time sequence offset matrix and a light source energy compensation matrix, constructing a light source collaborative error model through matrix superposition and eigenvalue decomposition, inputting exposure disturbance obtained by real-time monitoring of a photoetching machine into the error model, and obtaining a primary compensation amount; a final target exposure compensation parameter is obtained by adopting a quadratic optimization algorithm under a constraint condition, and light source energy output, a light path phase and a pulse trigger time sequence are cooperatively adjusted according to the target exposure compensation parameter, so that energy deviation, phase drift and time sequence mismatch can be reduced at the same time, the exposure stability and the pattern forming precision are remarkably improved, and the image forming precision is improved. The method is suitable for high-resolution and multi-light-source photoetching equipment.

This application relates to the field of scanner image acquisition technology, and in particular to a method and apparatus for dynamic compensation of exposure time of a reverse-side CIS sensor. The method is applied to a duplex automatic document feeder scanning system, which is equipped with independent CIS sensors for the front and back sides to achieve synchronous scanning of the front and back sides. The method includes: upon receiving a duplex scanning task instruction initiated by a user, obtaining the current relative luminous intensity of the reverse-side CIS sensor LED based on the current cumulative illumination time of the LED and a pre-stored attenuation curve model of the reverse-side CIS sensor LED; obtaining the compensated exposure time based on the current relative luminous intensity and pre-stored initial parameters of the reverse-side CIS sensor; and controlling the reverse-side CIS sensor LED to illuminate according to the compensated exposure time, thereby achieving exposure compensation and image quality maintenance of the reverse-side image during duplex scanning.

The invention discloses a self-adaptive fingerprint identification method and module based on multi-modal imaging, and the method comprises the steps: obtaining a multi-spectral fingerprint image sequence, triggering a plurality of light sources in a time-sharing manner, and collecting a spectrum corresponding to each light source; acquiring a stain type and a regional mask image, inputting the multispectral fingerprint image sequence into a target light convolutional neural network model, and performing stain type identification by the target light convolutional neural network model according to the multispectral fingerprint image sequence; and performing adaptive fingerprint identification, performing image enhancement according to the stain type to obtain a first enhanced image, performing local dynamic exposure compensation on the first enhanced image to obtain a second enhanced image, and transmitting the second enhanced image to a fingerprint identification unit for identification. The fingerprint recognition accuracy is improved, and the problems that the effect is poor when a traditional enhanced image software algorithm is used for recognition, and the fingerprint recognition accuracy is low due to the fact that dynamic stain type judgment is not combined are solved.

The invention provides an image processing method and electronic equipment. The problem that an image shot by the electronic equipment has color cast can be solved. The method comprises the steps of displaying a first image in response to a shooting operation of a user; the first image is generated by fusing a second long exposure image frame and a second short exposure image frame; the exposure value of the second long-exposure image frame is a third exposure value, and the second short-exposure image frame is obtained after exposure compensation is carried out by adopting the target gray-scale value; wherein the third exposure value is obtained by adopting an automatic exposure AE algorithm, and the target gray-scale value is obtained based on the relationship between the gray-scale value of the image frame and the number of pixels.

The invention relates to a marine unstructured camera video stitching method, which comprises the following steps: S1, acquiring video stream data of a shipborne camera, the video stream data being initial data required for calculating stitching parameters; s2, the initial data obtained in the step 1 are processed, splicing mapping parameters are obtained, and the splicing mapping parameters comprise cylindrical projection mapping parameters, homography matrixes between adjacent cameras, self exposure compensation parameters and optimal suture lines; splicing mapping parameters are stored in a file form, and the parameters are loaded to a memory in a GPU mapping table form during subsequent splicing; and S3, on the basis of the splicing mapping parameters obtained in the step S2, carrying out online splicing on the video images. According to the method, the problem that the ship camera array is difficult to calibrate is solved, a more stable splicing effect is achieved in different illumination scenes, and compared with a traditional video splicing method, the method is higher in dynamic scene adaptability.

The invention discloses a semantic 3D Gaussian sputtering SLAM (Simultaneous Localization and Mapping) method based on illumination invariance. The method comprises the following steps: step 1, performing 3D Gaussian-based scene representation and joint micro-rendering on a to-be-constructed target area to obtain a multi-dimensional rendered initial scene model; 2, an internal appearance normalization IAN module is constructed, and decoupling of internal attributes of an initial scene model and transient illumination is achieved; 3, performing offset correction on a radiation field of a rendering graph in the initial scene model, and designing DRB-Loss loss to dynamically adjust correction intensity; step 4, through camera tracking optimization, combining a stable map and exposure compensation to realize robust pose estimation, and through double-stage key frame screening, improving SLAM system efficiency; step 5, constructing joint map loss, and optimizing all parameters of the 3D Gaussian primitive; and map updating is completed, invalid 3D Gaussian primitives are pruned, and a complete synchronous positioning and map building cycle is formed.

The application discloses a photolithography machine multi-light-source cooperative high-precision exposure control system and relates to the technical field of photolithography machine control. The initial energy parameters of each light source are acquired, a coupling characteristic set of energy variation and light spot appearance variation is constructed, light intensity error, phase error and light spot distortion error are taken as inputs, a time sequence offset matrix between the light sources and a light source energy compensation matrix are respectively generated, a light source cooperative error model is constructed through matrix superposition and eigenvalue decomposition, exposure disturbance obtained by real-time monitoring of the photolithography machine is input into the error model, the final target exposure compensation parameters are obtained by using a quadratic optimization algorithm under a constraint condition after the preliminary compensation amount is obtained, the light source energy output, the light path phase and the pulse trigger time sequence are cooperatively adjusted according to the target exposure compensation parameters, the energy deviation, the phase drift and the time sequence mismatch can be simultaneously reduced, the exposure stability and the pattern shaping precision are significantly improved, and the application is suitable for high-resolution and multi-light-source photolithography equipment.

The invention relates to the technical field of image processing, in particular to an MLCC metallographic microscopic image splicing method and system for repeated textures, and the method comprises the following steps: S1, obtaining a first image and a second image; s2, carrying out exhaustion matching by adopting a violent matcher to obtain an initial matching point pair set; s3, performing mismatching elimination processing on the initial matching point pair set to obtain a final inner point matching set; s4, calculating an affine transformation matrix between the first image and the second image; s5, aligning images in the image sequence in the affine transformation matrix by adopting a chain accumulation transformation mode to obtain a third image, and performing exposure compensation processing on the third image to obtain a fourth image; s6, an optimal path search algorithm is adopted to determine a splicing seam with the minimum visual cost in the overlapping area of the fourth image, feathering fusion is carried out along the splicing seam, and a fifth image is obtained; s7, cutting the fifth image to obtain a target spliced image; and the problem of matching fuzziness caused by repeated textures is effectively solved.

The invention relates to the technical field of crossing of artificial intelligence and machine vision, discloses an AI-vision-based machine tool frame casting defect detection method and system, and aims to solve the technical problems that the image quality is reduced and the defect detection accuracy is influenced due to sudden change of instantaneous illumination in an industrial field. According to the method, illumination time sequence data are collected in real time by integrating a main imaging sensor and a high-frequency auxiliary light field sensor array, and future light field distribution is predicted based on a hardware Kalman filter; and generating a pixel-level dynamic gain matrix according to the main image, performing regional adaptive exposure compensation on the main image in a simulation domain, performing linear reconstruction and gain adaptive denoising, and outputting an illumination balanced image for an AI model to detect microcracks and pores. The system realizes microsecond-level illumination abrupt change response and local accurate compensation, and obviously improves the imaging consistency and defect detection rate.

The invention relates to the technical field of exposure control, in particular to a tea garden camera dynamic exposure control method and system based on a random forest prediction model, and the method comprises the steps: S1, collecting image data of a tea garden under different illumination conditions, and extracting image quality parameters of a tender shoot region; s2, synchronously acquiring ambient light parameters when each image is shot; s3, constructing and training a random forest prediction model; s4, acquiring a current ambient light parameter and an image quality parameter of a previous frame of image in real time; s5, inputting the current ambient light parameter and the image quality parameter of the previous frame of image into the prediction model, and predicting the optimal exposure parameter of the current frame; and S6, adjusting camera exposure setting according to the optimal exposure parameter, and collecting a current frame image. And S7, collecting the local light-induced response difference spectrum, performing comparison to generate an exposure compensation sensitivity spectrum, and performing local offset correction. The image definition is improved, and the method is suitable for an image acquisition task highly sensitive to exposure quality.

The invention provides a non-uniform illumination underground mine scene reconstruction method based on three-dimensional Gaussian sputtering. The method comprises the following steps: firstly, acquiring mine vision data, processing to generate a multi-view two-dimensional image, constructing a sparse point cloud, and initializing a three-dimensional Gaussian ellipsoid set containing dynamic structure features to obtain an initial three-dimensional Gaussian body; and extracting a mine physical prior structure diagram as a geometric growth supervision signal. Performing decoupling rendering on the initial Gaussian body to generate a vector chromaticity diagram, a physical radiation diagram and a structural feature diagram; and performing exposure compensation on the physical radiation pattern by combining camera parameters, inhibiting dark part noise through an intensity collapse function to obtain a clean physical intensity pattern, recombining the clean physical intensity pattern into an HSV component and inversely transforming the HSV component into an RGB image. And finally, constructing a multi-loss constraint based on the RGB image, the structural feature graph, the physical prior structure graph and a real image, performing end-to-end training by adopting an Adam algorithm to update Gaussian parameters to convergence, and outputting a three-dimensional reconstruction result through visual angle adaptive rendering, thereby giving consideration to robustness, high precision and real-time performance.

The application relates to the technical field of image processing, and provides an image fusion method and device, a storage medium and an electronic device. The image fusion method comprises the following steps: acquiring a plurality of first-type images, the exposure compensation values of the first-type images are all different, and the exposure compensation values of at least two of the first-type images are non-positive numbers; calculating a first-type weight image between each two adjacent first-type images; calculating a highlight image corresponding to each first-type image with a non-positive exposure compensation value; calculating a real highlight image containing a highlight region and / or a white region according to each two adjacent highlight images; correcting a weight value in the corresponding first-type weight image by using the real highlight image; and performing weighted fusion on the plurality of first-type images by using at least the corrected first-type weight image, so as to obtain a final fusion image. The method can effectively improve the problem of grayness in the high-luminance region or the white region of the fusion image, and improve the quality of the fusion image.

The invention relates to the technical field of photoetching machine optical regulation and control, and discloses an intelligent feedback alignment adjustment micro-control system for an MLA lens of a photoetching machine. The system comprises a light intensity distribution acquisition module, a lens state mapping module, an offset optimization module, an exposure compensation decision module and a dynamic calibration feedback module. The light intensity distribution acquisition module acquires light intensity distribution data in an exposure field in real time, and analyzes spatial characteristics and a time change sequence to generate an extraction result; the lens state mapping module is associated with physical coordinates of the lens units, and optical parameter offset is calculated to generate a mapping set; the offset optimization module analyzes the accumulated distribution and fluctuation range of the offset, and generates a multi-dimensional optimization parameter set; the exposure compensation decision module dynamically distributes compensation priorities, and calculates compensation coefficients and fine tuning vectors to generate an execution instruction set; and the dynamic calibration feedback module executes the instruction and monitors and feeds back calibration deviation. The system realizes accurate adjustment of lens alignment, and ensures the exposure quality.

The invention discloses a visual guidance control method for an underwater hull cleaning robot, and relates to the technical field of visual guidance, and the method comprises the steps: obtaining underwater original images of the vertical and inclined surfaces of a broadside hull, and carrying out the longitudinal partitioning of the images along the broadside height based on the gravity direction; in combination with partition brightness distribution prior, adaptive exposure compensation is performed on each partition, brightness mutation of adjacent partitions is eliminated through longitudinal consistency constraint, and a broadside image with continuous brightness is generated; on this basis, hull weld joint and reinforcing rib structure feature lines are extracted, a visual reference coordinate system consistent with the hull structure is constructed, then real-time deviation correction control is conducted on the cleaning robot according to the visual offset, and the cleaning robot is guided to execute a stable and continuous broadside cleaning path in the direction of the structure feature lines; the problems of visual distortion, unstable structural feature recognition and easy deviation of a cleaning path caused by non-uniform illumination and posture change in an underwater broadside cleaning scene are solved.

The invention relates to the related technical field of image recognition analysis, in particular to a sesame quality detection method and system based on image recognition, and the method comprises the steps: obtaining sesame epidermis and endosperm spectral data and microscopic imaging features, setting a texture matrix, extracting morphological features, finely tuning a pre-training model for multi-dimensional fusion, and dynamically calibrating exposure time and light source intensity. And performing exposure compensation control when the environment changes. The technical problems of unstable image acquisition quality, difficulty in accurately distinguishing tiny defects and limited sesame quality detection precision caused by ambient light intensity fluctuation and sesame stacking density change are solved, and the purpose of accurately detecting the sesame quality by synchronously acquiring reflection spectrum data of an epidermal wax coat and internal endosperm and dynamically calibrating exposure time and light source intensity in combination with a microscopic texture feature matrix is achieved. Interference of ambient light intensity fluctuation and sesame stacking density change is effectively resisted, weight distribution is performed on form, color and texture features in combination with an attention mechanism, classification boundaries are optimized, and the technical effects of sesame quality detection precision are guaranteed.

The application discloses a focusing calibration method and an exposure compensation method with an automatic focusing function, determines the relationship between the movement amount of an AF mechanism and the change amount of a platform focal plane, and specifically comprises the following steps: arranging the AF mechanism below an exposure lens, moving a suction disc camera on the platform to be directly below the exposure lens, determining the optimal focal plane position Z of the platform through the suction disc camera m ; determining the relationship between the movement amount of the AF mechanism and the change amount of the platform focal plane, obtaining the relationship model between the focusing position of the AF mechanism and the change amount of the platform focal plane, thereby ensuring the accuracy of the focal plane compensation of the AF mechanism, eliminating the different influences of machining errors and movement errors and movement amounts on the focal plane, and preventing the phenomenon that the difference between the theoretical compensation surface and the actual compensation surface is different.

The invention provides a garage fire detection method and system based on dynamic exposure compensation, and the method comprises the steps: obtaining a to-be-detected image from a garage video stream, analyzing a brightness image of the to-be-detected image, and calculating a dynamic gain and a dynamic bias; linear transformation is carried out on the RGB image according to the dynamic gain and the dynamic bias, and dynamic exposure compensation is completed; the method comprises the following steps: firstly, training an image to be detected, then, carrying out target detection according to a trained yolov8 network, identifying whether a target with suspected flame exists in the image to be detected, if so, carrying out labeling and generating regional information, finally, calculating spectral distribution and brightness variance of a fire occurrence region, carrying out reconfirmation, and finally, giving a fire detection result. And through multi-aspect combined judgment of spectral characteristics, the fire detection accuracy is improved, and the frequency of misjudgment and missed judgment is reduced.

The application provides a mobile robot binocular vision global positioning system and method, relates to the technical field of robot vision navigation, and extracts light-independent edge contour feature points by performing adaptive exposure compensation on binocular images of left and right cameras, combines with parallax calculation to generate a high-quality depth feature map, simultaneously constructs a feature unit according to an inflection point, and thus the global position of the mobile robot is determined. In this way, the problem that the global positioning accuracy is reduced due to unstable vision feature extraction caused by sharp changes in light intensity in the multi-region switching scene of the mobile robot can be solved to a certain extent.

The application provides an image anti-shake control method, an electronic device, a chip system and a storage medium, and relates to the technical field of terminals. In a shooting scene of a camera night scene mode, a first anti-shake control parameter suitable for a current scene can be determined based on scene brightness information, such as a shake information filtering mode, an image frame output mode and / or an image fusion mode, and a second anti-shake control parameter suitable for the current scene can be determined based on shake information, such as an exposure compensation amount, different shake information corresponding to different exposure compensation amounts. The application scheme can perform exposure compensation of different degrees in the shooting stage for night scene shooting scenes with different shake intensities, realize the image anti-shake purpose, and realize comprehensive anti-shake control and image processing in combination with the shake information filtering mode, the image frame output mode and / or the image fusion mode, thereby improving the image anti-shake effect of the night scene shooting scene and improving the definition of the night scene photo.

The embodiment of the invention provides an exposure adjustment method, device and system and a computer readable storage medium, and relates to the field of single crystal drawing, the method comprises the following steps: determining an initial gray value of an image acquired by imaging equipment based on initial exposure; under the condition of entering the target stage, determining a real-time gray value of an image acquired by the imaging equipment based on the initial exposure; determining real-time exposure based on the real-time gray value, the initial gray value, the initial exposure and the target gray value; and adjusting the exposure of the imaging equipment to be the real-time exposure. According to the method, self-adaptive exposure compensation of the imaging equipment can be realized, so that the brightness in the furnace can be compensated during imaging, the light environment of a Czochralski single crystal re-casting step and a material melting step is improved, the image acquisition quality is improved, and the accuracy of silicon material state detection according to the acquired image is further improved.

The invention relates to the field of medical instruments, and discloses a fluorescence exposure compensation method, system and device for a 3D fluorescence endoscope camera system, and the method comprises the following steps: S1, parameter calibration: before a 3D fluorescence endoscope operation is carried out, calibration of key parameters of the system needs to be completed, and an accurate reference basis is provided for real-time fluorescence exposure compensation in the operation; s2, fluorescent camera exposure parameter estimation: in the 3D fluorescent endoscopic surgery process, estimating and adjusting the fluorescent camera exposure parameters in real time based on the parameters calibrated in S1, so as to realize stable fluorescent exposure. According to the method, the relation curve of the shooting distance and the parallax is established through preoperative dual-target calibration, the parallax image is estimated based on the binocular visible light image and the weighted average parallax value is calculated in the operation, and compared with an existing distance estimation mode based on white light exposure parameters and brightness, the estimation precision of the shooting distance is greatly improved, and the accuracy of shooting distance estimation is improved. And an accurate distance basis is provided for fluorescence exposure compensation.

The invention discloses a multi-image panoramic stitching method based on depth matching, a camera and a storage medium, and the method comprises the following steps: S1, image preprocessing and multi-scale zooming: firstly inputting a multi-image sequence, and after inputting the multi-image sequence, according to a self-defined working resolution, a seam resolution and a synthesis resolution, carrying out the multi-scale zooming of the multi-image sequence; the method comprises the following steps of: performing adaptive scaling on a plurality of input images to generate a multi-scale image set; s2, deep learning end-to-end matching: performing end-to-end feature matching on the multi-scale image set by using a pre-trained ONNX model to generate direct matching information between the image pairs; s3, performing global consistency screening; s4, estimating and optimizing camera parameters; and S5, performing exposure compensation and image fusion. The method has the advantages that the matching precision and the calculation efficiency are remarkably improved, the overall consistency of the splicing result is ensured, the calculation efficiency and the splicing quality are balanced, and the method is suitable for ultrahigh-resolution image sequences.

The invention relates to the technical field of computer vision and unmanned aerial vehicle inspection, in particular to a target area photometry and exposure mapping method and system in transmission tower inspection. The method comprises the following steps: determining a target area of a pole and tower key component and an externally-expanded annular belt background area, calculating a target brightness statistic and a highlight risk index in the target area, and calculating an exposure compensation amount in combination with a background brightness statistic; under the constraint of the upper limit of the shutter time and the upper limit of the sensitivity, quickly mapping the exposure compensation amount into target shutter time, target sensitivity and a target exposure compensation value; step limitation, exponential smoothing and hysteresis processing are carried out on the mapping parameters, and issuing exposure parameters are obtained; and when the exposure is stable and the image quality meets the condition, outputting a photographing trigger signal to form continuous closed-loop control. According to the method, the problems of uncontrollable exposure, slow convergence and easy oscillation in a strong-reflection and high-dynamic-range scene of a traditional method are solved, and the inspection image quality and the defect identification reliability are improved.